Paramedic Papers Daily — 8 November 2026

Four papers today, all cardiac. Paper one — mainstream and sidestream capnography give systematically different end-tidal carbon dioxide numbers during CPR, but the trend is the same. Paper two — Swedish community first responders feel well prepared, yet almost none have ever actually been dispatched. Paper three — a scoping review finds eight chest-pain risk tools used in ambulance care, with HEART the most studied but validation still incomplete. And paper four — a policy analysis argues human oversight of AI in cardiac arrest care risks becoming more symbolic than real.

Paper one — Resuscitation Plus. A retrospective device-comparison study, twenty-three out-of-hospital cardiac arrest recordings with mainstream and sidestream capnography running at the same time — three thousand four hundred and sixty paired ventilations. Mainstream read systematically higher: a bias of nearly four millimetres of mercury, and the gap widened the higher the carbon dioxide climbed, with limits of agreement running from minus three to plus eleven. But the trends tracked each other almost identically — both technologies measured a falling end-tidal reading at essentially the same rate — and abrupt rises of ten or more were co-detected in seven of eight cases, drops in ten of fifteen. That matters because the absolute number a crew calls out at the hospital depends on which technology the monitor uses, and decision thresholds built around ten millimetres shouldn't be read blind to the device. The caveat: twenty-three recordings from one region, so those wide limits of agreement are imprecisely estimated.

Paper two — also Resuscitation Plus. A cross-sectional survey of community first responders across two Swedish regions, taken just before their local systems launched. Two thousand seven hundred and four responded out of more than fourteen thousand registered — a response rate of about nineteen per cent. Eighty-four per cent had CPR training within the last two years, and ninety-seven per cent felt prepared to respond. Yet only fifteen per cent had ever been dispatched, and just under two per cent had actually performed CPR. The readiness in this reservoir of volunteers is built on training rather than runs on the board — and it's self-reported, so take the confidence with a grain of salt. Community responder schemes operate in parts of Australia, and this is the same question any new scheme faces: enthusiasm and certification don't guarantee anyone shows up in time. The low response rate is worth remembering too.

Paper three — Paramedicine. A scoping review of twenty-seven articles mapping chest-pain risk-stratification tools in ambulance care. Eight tools have been used or trialled in ambulance settings, with the HEART score and its derivatives the most studied. But performance varied across settings, and validation specifically for ambulance use is incomplete. Chest pain is a big slice of the workload, and safely not transporting the low-risk patient is the prize every service is chasing — this review is a useful map of what's been tried, but it's descriptive only, with no pooled performance numbers. Most of these tools were born in the emergency department, and the prehospital evidence hasn't caught up.

Paper four — Resuscitation Plus again, and a different animal: a conceptual policy analysis with no data of its own. It asks whether human-in-the-loop oversight of AI in cardiac arrest care actually works. The argument: under time pressure, high cognitive load and protocol-driven practice, crews tend to defer to AI recommendations rather than genuinely appraise them, and offsite medical control gets undermined by communication friction and ambiguous authority. Human oversight, the authors say, risks becoming a governance fiction. They propose five fixes — trialling the systems in real-world conditions, explainability designed in from the start, clearly defined escalation triggers and authority, and protection for crews who deviate from an AI recommendation with good reason. Sensible points, but it's opinion — there's no evidence yet on how crews actually behave.

Quick recap. One — mainstream capnography reads systematically higher than sidestream during CPR, but trends and abrupt changes match. Two — Swedish community first responders are trained and confident, yet almost none have ever been dispatched. Three — eight chest-pain risk tools have reached ambulance care, led by HEART, but prehospital validation is incomplete. Four — a policy paper warns AI oversight in resuscitation may be symbolic, and sets out five safeguards.

That's the roundup for Sunday the eighth of November. Papers on the dashboard. Back tomorrow.